Several 3D imaging methods based on active illumina- tion, such as silhouette-based 3D reconstruc... more Several 3D imaging methods based on active illumina- tion, such as silhouette-based 3D reconstruction and struc- tured light 3D scanning with binary patterns, require light sources capable of generating shadows with sharp bound- aries when they are used to illuminate opaque occluders. Supported by empirical evidence suggesting that a low cost Uncollimated Laser Diode (ULD) produces shadows with sharp boundaries not requiring focusing in a wide range of depths, this paper proposes the use of ULDs as light sources in the target applications. Since due to astigmatism the Point Light Source (PLS) is not an accurate mathemati- cal model of light propagation for the ULD, the Two Lines Light Source (TLLS) model is introduced to explain the ob- served behavior of the ULD. This novel geometric model of light propagation is defined by two 3D line segments, rather than a single 3D point, and guarantees that for each illu- minated 3D point there exists a unique ray, which simul- taneously passes through the point and intersects the two line segments. Furthermore, the equation of this ray can be computed in closed form at very low computational cost, and the TLLS model reduces to the PLS model when the two line segments intersect. Finally, the paper introduces a calibration method to estimate the model parameters, and describes the experiments performed to validate the model.
2021 International Conference on 3D Vision (3DV), 2021
Digital pattern projectors used in structured light 3D scanners cannot project infinitesimally th... more Digital pattern projectors used in structured light 3D scanners cannot project infinitesimally thin columns of light – the projector pixel columns have a finite thickness. Current methods incorrectly model these columns of light as planes, which contributes to reconstruction errors. We propose a method to increase the resolution of structured light 3D scanners based on time-multiplexed discrete patterns by taking multiple acquisitions of the scene with different projector poses, and by modeling the light projected by each pixel column as a volume in space instead of as a plane, which results in upper and lower bounds on the resulting depth map. Furthermore, by analyzing multiple acquisitions in different projector poses, these upper and lower depth bounds can be tightened. We refer to this tightening as super-resolution, because it corresponds to an increase in the confidence in the location of the object’s surface. We describe our first implementation of such a system, and demonstrate its performance on a variety of objects.
We describe a project to create a three dimensional digital model of Michelangelo’s Florentine Pi... more We describe a project to create a three dimensional digital model of Michelangelo’s Florentine Pietà. In particular, we examine how to edit and interactively view the model that consists of millions of tri angles and hundreds of color and normal maps.
2018 International Conference on Computing Sciences and Engineering (ICCSE), 2018
In this paper, we introduce a new method for the rough alignment of point-clouds. We focus on a s... more In this paper, we introduce a new method for the rough alignment of point-clouds. We focus on a special type of point-clouds that is composed of simple geometric shapes like planes, cylinders, cones, etc. We call them 3D CAD point clouds. They are usually used in industrial and mechanical applications. The proposed method starts by detecting basic shapes in the point-clouds. And then using them to find the best transformation (rotation and translation) that aligns the point- clouds. Then, we run the fine alignment step using the iterative closest point method (ICP). We show several real-world examples of point-clouds before and after the alignment using this method. The results suggest that the proposed method works well in most cases given enough overlap between the point-clouds.
2015 IEEE International Conference on Image Processing (ICIP), 2015
Image inpainting is a challenging topic in computer vision that seeks to recover the natural aspe... more Image inpainting is a challenging topic in computer vision that seeks to recover the natural aspect of an image where data has been partially damaged or occluded by undesired objects. A common drawback not addressed by most inpainting methodologies is that the user must manually provide the inpainting mask as input data to the method. Selecting the inpainting mask is tedious, time consuming and it often requires artistic skills to precisely determine the mask. In this work we design a new tool that allows users to easily select the desirable mask. The proposed framework combines the high-adherence on image contours of the Laplacian Coordinates segmentation approach with the efficiency of a recent inpainting technique that unifies anisotropic diffusion, inner product-based filling order mechanism and exemplar-based completion. The user can interact with the object that he/she intends to edit by stroking small parts of the object so as to proceed with the segmentation and inpainting task. Our comparisons show that the proposed framework has good performance in terms of applicability and effectiveness when compared against other existing techniques in the literature.
The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household ... more The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household items and a camera. Specifically, we’ll implement the “desktop scanner” originally proposed by Jean-Yves Bouguet and Pietro Perona [3]. As shown in Figure 1, our instantiation of this system is composed of five primary items: a camera, a point-like light source, a stick, two planar surfaces, and a calibration checkerboard. By waving the stick in front of the light source, the user can cast planar shadows into the scene. As we’ll demonstrate in this handout, the depth at each pixel can then be recovered using simple geometric reasoning. In the course of completing this homework, you will need to develop a good understanding of camera calibration, Euclidean coordinate transformations, manipulation of implicit and parametric parameterizations of lines and planes, and efficient numerical methods for solving least-squares problems. Before you begin this assignment, you should first browse the...
The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household ... more The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household items and a camera. Specifically, we’ll implement the “desktop scanner” originally proposed by Jean-Yves Bouguet and Pietro Perona [3]. As shown in Figure 1, our instantiation of this system is composed of five primary items: a camera, a point-like light source, a stick, two planar surfaces, and a calibration checkerboard. By waving the stick in front of the light source, the user can cast planar shadows into the scene. As we’ll demonstrate in this handout, the depth at each pixel can then be recovered using simple geometric reasoning. In the course of completing this homework, you will need to develop a good understanding of camera calibration, Euclidean coordinate transformations, manipulation of implicit and parametric parameterizations of lines and planes, and efficient numerical methods for solving least-squares problems. Before you begin this assignment, you should first browse the...
Several 3D imaging methods based on active illumina- tion, such as silhouette-based 3D reconstruc... more Several 3D imaging methods based on active illumina- tion, such as silhouette-based 3D reconstruction and struc- tured light 3D scanning with binary patterns, require light sources capable of generating shadows with sharp bound- aries when they are used to illuminate opaque occluders. Supported by empirical evidence suggesting that a low cost Uncollimated Laser Diode (ULD) produces shadows with sharp boundaries not requiring focusing in a wide range of depths, this paper proposes the use of ULDs as light sources in the target applications. Since due to astigmatism the Point Light Source (PLS) is not an accurate mathemati- cal model of light propagation for the ULD, the Two Lines Light Source (TLLS) model is introduced to explain the ob- served behavior of the ULD. This novel geometric model of light propagation is defined by two 3D line segments, rather than a single 3D point, and guarantees that for each illu- minated 3D point there exists a unique ray, which simul- taneously passes through the point and intersects the two line segments. Furthermore, the equation of this ray can be computed in closed form at very low computational cost, and the TLLS model reduces to the PLS model when the two line segments intersect. Finally, the paper introduces a calibration method to estimate the model parameters, and describes the experiments performed to validate the model.
2021 International Conference on 3D Vision (3DV), 2021
Digital pattern projectors used in structured light 3D scanners cannot project infinitesimally th... more Digital pattern projectors used in structured light 3D scanners cannot project infinitesimally thin columns of light – the projector pixel columns have a finite thickness. Current methods incorrectly model these columns of light as planes, which contributes to reconstruction errors. We propose a method to increase the resolution of structured light 3D scanners based on time-multiplexed discrete patterns by taking multiple acquisitions of the scene with different projector poses, and by modeling the light projected by each pixel column as a volume in space instead of as a plane, which results in upper and lower bounds on the resulting depth map. Furthermore, by analyzing multiple acquisitions in different projector poses, these upper and lower depth bounds can be tightened. We refer to this tightening as super-resolution, because it corresponds to an increase in the confidence in the location of the object’s surface. We describe our first implementation of such a system, and demonstrate its performance on a variety of objects.
We describe a project to create a three dimensional digital model of Michelangelo’s Florentine Pi... more We describe a project to create a three dimensional digital model of Michelangelo’s Florentine Pietà. In particular, we examine how to edit and interactively view the model that consists of millions of tri angles and hundreds of color and normal maps.
2018 International Conference on Computing Sciences and Engineering (ICCSE), 2018
In this paper, we introduce a new method for the rough alignment of point-clouds. We focus on a s... more In this paper, we introduce a new method for the rough alignment of point-clouds. We focus on a special type of point-clouds that is composed of simple geometric shapes like planes, cylinders, cones, etc. We call them 3D CAD point clouds. They are usually used in industrial and mechanical applications. The proposed method starts by detecting basic shapes in the point-clouds. And then using them to find the best transformation (rotation and translation) that aligns the point- clouds. Then, we run the fine alignment step using the iterative closest point method (ICP). We show several real-world examples of point-clouds before and after the alignment using this method. The results suggest that the proposed method works well in most cases given enough overlap between the point-clouds.
2015 IEEE International Conference on Image Processing (ICIP), 2015
Image inpainting is a challenging topic in computer vision that seeks to recover the natural aspe... more Image inpainting is a challenging topic in computer vision that seeks to recover the natural aspect of an image where data has been partially damaged or occluded by undesired objects. A common drawback not addressed by most inpainting methodologies is that the user must manually provide the inpainting mask as input data to the method. Selecting the inpainting mask is tedious, time consuming and it often requires artistic skills to precisely determine the mask. In this work we design a new tool that allows users to easily select the desirable mask. The proposed framework combines the high-adherence on image contours of the Laplacian Coordinates segmentation approach with the efficiency of a recent inpainting technique that unifies anisotropic diffusion, inner product-based filling order mechanism and exemplar-based completion. The user can interact with the object that he/she intends to edit by stroking small parts of the object so as to proceed with the segmentation and inpainting task. Our comparisons show that the proposed framework has good performance in terms of applicability and effectiveness when compared against other existing techniques in the literature.
The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household ... more The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household items and a camera. Specifically, we’ll implement the “desktop scanner” originally proposed by Jean-Yves Bouguet and Pietro Perona [3]. As shown in Figure 1, our instantiation of this system is composed of five primary items: a camera, a point-like light source, a stick, two planar surfaces, and a calibration checkerboard. By waving the stick in front of the light source, the user can cast planar shadows into the scene. As we’ll demonstrate in this handout, the depth at each pixel can then be recovered using simple geometric reasoning. In the course of completing this homework, you will need to develop a good understanding of camera calibration, Euclidean coordinate transformations, manipulation of implicit and parametric parameterizations of lines and planes, and efficient numerical methods for solving least-squares problems. Before you begin this assignment, you should first browse the...
The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household ... more The goal of this assignment is to build an inexpensive, yet accurate, 3D scanner using household items and a camera. Specifically, we’ll implement the “desktop scanner” originally proposed by Jean-Yves Bouguet and Pietro Perona [3]. As shown in Figure 1, our instantiation of this system is composed of five primary items: a camera, a point-like light source, a stick, two planar surfaces, and a calibration checkerboard. By waving the stick in front of the light source, the user can cast planar shadows into the scene. As we’ll demonstrate in this handout, the depth at each pixel can then be recovered using simple geometric reasoning. In the course of completing this homework, you will need to develop a good understanding of camera calibration, Euclidean coordinate transformations, manipulation of implicit and parametric parameterizations of lines and planes, and efficient numerical methods for solving least-squares problems. Before you begin this assignment, you should first browse the...
Uploads